Multi-model based family and personal account relationship identification method and system

Abstract

The present invention relates to the field of Internet technology and provides a method and system for identifying relationships between household and personal accounts based on multiple models. The method comprises: obtaining at least two sets of association results for the same household account based on different identification models; calculating the association credibility between the household account and the personal account corresponding to the household account based on the at least two sets of association results and the model confidence of the identification model; and adding the household account and the personal account corresponding to the household account whose association credibility meets a set range to the identification conclusion. The present invention obtains association results through multiple models and verifies the association credibility of the household account and the personal account based on the model confidence. This improves the integrity of the association results through multi-angle data from multiple models and increases the accuracy of the association results through cross-validation integration based on association credibility.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a method and system for identifying the relationship between family and personal accounts based on multiple models. Background Art

[0002] Telecommunications operators offer multiple CHBN services, including personal services, family services, enterprise services, and new services. Personal and family services have overlapping customer bases. Based on the relationship between family and personal communication accounts, operators conduct correlation analysis for different services for the same customer, improving user satisfaction, user profiling, and targeted marketing efforts.

[0003] Typical personal services include mobile communications services such as data and internet access, VoLTE, and other services, using IMSI or MSISDN (mobile phone number) as the unique user identifier. Typical home services include home broadband, IPTV, and internet TV, using the home broadband account as the unique user identifier.

[0004] The current method for operators to obtain the association between family accounts and personal accounts is to associate and extract the personal accounts (such as mobile phone numbers) registered and retained by users in the subscription data of family account services (such as broadband services).

[0005] The above-mentioned prior art has the following problems:

[0006] 1. The correlation relationship is inaccurate.

[0007] Take the mobile phone number you saved when applying for home broadband service as an example:

[0008] First, the person who registers for a home broadband plan isn't necessarily the person who uses it. There are cases where a child registers and a parent uses it, or a landlord registers and a tenant uses it. In these cases, the phone number recorded isn't the actual user of the home broadband plan.

[0009] Second, retained mobile numbers may not be primary numbers or may belong to different carriers. With the prevalence of dual-SIM phones and declining rates, more and more users have multiple SIM cards from different carriers. Service analysis of retained non-primary numbers cannot reflect the user's true mobile service characteristics and perceptions. For those who retain mobile numbers from other carriers, their mobile and home services belong to different carriers and cannot be analyzed in a linked manner.

[0010] 2. The association relationship is incomplete.

[0011] There are usually multiple individual users in the same residence, such as different family members. Using the contract data method, a family account can only be linked to one individual account.

[0012] Therefore, how to more accurately identify the relationship between family and personal accounts has become a technical problem that urgently needs to be solved in the industry. Summary of the Invention

[0013] The present invention provides a method and system for identifying the relationship between family and personal accounts based on multiple models, which is used to solve the defects of inaccurate and incomplete association relationships in the existing technology and achieve more accurate identification of the relationship between family and personal accounts.

[0014] The present invention provides a method for identifying the relationship between household and personal accounts based on multiple models, comprising:

[0015] For the same family account, at least two sets of association results are obtained based on different recognition models; the association results are the results of the family account and the individual accounts corresponding to the family account in the association result set; the association result set is the set of correspondences between multiple family accounts and multiple individual accounts obtained by the recognition model based on the input data set;

[0016] Calculating, based on the at least two sets of association results and the model confidence of the recognition model, an association credibility between the family account and the personal account corresponding to the family account;

[0017] The family account whose association credibility meets the set range and the personal account corresponding to the family account are added to the identification conclusion.

[0018] According to a multi-model based method for identifying the relationship between household and personal accounts provided by the present invention, the identification model includes a baseline identification model and a comparative identification model; the model confidence of the baseline identification model is a set baseline confidence; the model confidence of the comparative identification model is a relative confidence obtained based on the association result set of the baseline identification model, the association result set of the comparative identification model and the baseline confidence.

[0019] According to a multi-model based household and personal account relationship identification method provided by the present invention, the relative confidence C of the comparative identification model satisfies:

[0020]

[0021] Where C0 is the baseline confidence level; n is the number of accounts in the first household; m is the number of accounts in the second household;

[0022] The first family account refers to a family account in the association result set of the comparative identification model that belongs to the association result set of the benchmark identification model; the second family account refers to the first family account including at least one correct personal account; the correct personal account refers to a personal account corresponding to the family account that is the same as the first family account in the association result set of the benchmark identification model.

[0023] According to a multi-model-based method for identifying relationships between family and personal accounts provided by the present invention, the step of calculating the association credibility between the family account and the personal account corresponding to the family account based on the at least two sets of association results and the model confidence of the identification model includes:

[0024] Based on the association result, determine whether the personal accounts corresponding to the family account include the target personal account:

[0025] If the target personal account is included, the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result;

[0026] If the target personal account is not included and the family account does not have a corresponding personal account, the credibility factor of the association result is set to zero;

[0027] If the target personal account is not included, and the family account has a corresponding personal account, the inverse of the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result;

[0028] The credibility factors of at least two groups of association results are superimposed to obtain the association credibility between the family account and the target personal account.

[0029] According to a multi-model based method for identifying the relationship between household and personal accounts provided by the present invention, the identification model includes any one or a combination of any two of a resident location elimination model based on business support data, a personal account mining model based on household account communications, a user relationship mining model based on application user identification, a relationship identification model based on business complaint data, a relationship identification model based on installation and maintenance data, and a relationship identification model based on account service application data.

[0030] According to a multi-model-based method for identifying relationships between family and personal accounts provided by the present invention, the business support data-based resident location elimination model is based on the relationship data between family and personal accounts in the business support data, the nighttime resident location of the personal account, and the range of the family account's resident network cell. The model eliminates association results in the business support data between family and personal accounts where the nighttime resident location of the personal account is not within the range of the family account's resident network cell.

[0031] The nighttime permanent location of the personal account is obtained through control plane signaling.

[0032] According to a multi-model based method for identifying the relationship between family and personal accounts provided by the present invention, the personal account mining model based on family account communication obtains the device identification code of the communication device based on the family account communication data, and obtains the model of the personal account used by the communication device based on the device identification code.

[0033] According to a multi-model based method for identifying the relationship between family and personal accounts provided by the present invention, the user relationship mining model based on application user identification is a model for obtaining family accounts and personal accounts that use the same user identification to communicate based on application user identification data.

[0034] The present invention also provides a multi-model-based family and personal account relationship identification system, comprising:

[0035] An association module, configured to obtain at least two sets of association results for the same family account based on different recognition models; the association results are included in a set of association results, including the family account and the individual accounts corresponding to the family account; the association result set is obtained by the recognition model based on an input data set, and includes a set of correspondences between multiple family accounts and multiple individual accounts;

[0036] a trust module, configured to calculate, based on the at least two sets of association results and the model confidence of the recognition model, an association credibility between the family account and the personal account corresponding to the family account;

[0037] The conclusion module is used to add the family account whose association credibility meets the set range and the personal account corresponding to the family account to the identification conclusion.

[0038] The present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of any of the above-described methods for identifying the relationship between household and personal accounts based on multiple models are implemented.

[0039] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the multi-model-based methods for identifying the relationship between household and personal accounts as described above.

[0040] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of any of the above-mentioned methods for identifying the relationship between household and personal accounts based on multiple models.

[0041] The multi-model-based family and personal account relationship identification method and system provided by the present invention obtains association results through multiple models, and verifies the association results based on the association credibility of family accounts and personal accounts based on the model confidence. On the one hand, the integrity of the association results is improved through the multi-angle data of multiple models, and on the other hand, the accuracy of the association results is higher through cross-validation integration based on association credibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 1 is a flow chart of a multi-model based household and personal account relationship identification method provided by the present invention;

[0044] Figure 2 Schematic diagram of the structure of the multi-model based family and personal account relationship identification system provided by the present invention;

[0045] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0047] The following combination Figure 1 The present invention describes a multi-model-based method for identifying relationships between household and personal accounts.

[0048] like Figure 1 As shown, an embodiment of the present invention provides a method for identifying the relationship between household and personal accounts based on multiple models, including:

[0049] Step 101: Obtain at least two sets of association results for the same family account based on different recognition models; the association results are the results of the family account and the individual accounts corresponding to the family account in the association result set; the association result set is a set of correspondences between multiple family accounts and multiple individual accounts, obtained by the recognition model based on the input data set;

[0050] Step 103: Calculate the association credibility between the family account and the personal account corresponding to the family account based on the at least two sets of association results and the model confidence of the recognition model;

[0051] Step 105: Add the family account whose association credibility meets the set range and the personal account corresponding to the family account to the identification conclusion.

[0052] In step 101 of this embodiment, the association result refers to the matching relationship between the family account and the individual account. In some embodiments, this can be understood as meaning that the individual account holder is the user of the family account service, and is not limited to the individual account holder being a family member or having a family relationship. Furthermore, in the association result, a family account can correspond to zero, one, or multiple individual accounts.

[0053] In some preferred embodiments, each step of this embodiment can be understood as a verification and screening process for a single specific family account. After repeating the steps of this embodiment to traverse all possible family accounts, a complete data set of correspondences between family accounts and personal accounts, i.e., an identification conclusion, can be obtained.

[0054] In this embodiment, the input data set of the recognition model may be updated based on a set time interval, and thus the association result and the recognition conclusion may also be updated based on a set time interval.

[0055] In this embodiment, acquiring a more comprehensive input data set over time can further improve the accuracy and completeness of the identification conclusions. However, considering the service period corresponding to the family accounts, data from family accounts that have exceeded their service period and have not renewed their services can be eliminated to further ensure the accuracy and completeness of the identification conclusions.

[0056] In a preferred embodiment, the elements contained in the set of identification conclusions are relationship data groups, which refer to family accounts and all personal accounts corresponding to the family accounts. That is, there are no duplicate family accounts in the identification conclusions, and all personal accounts that have a corresponding relationship with a specific family account are integrated into the relationship data group where the family account is located.

[0057] In another preferred embodiment, the elements included in the set of identification conclusions are relationship data pairs, which refer to corresponding relationship data pairs between family accounts and personal accounts that are consistent with the association results. In this case, there may be duplicate family accounts in the identification conclusions.

[0058] In addition, the user data involved in this embodiment and the following embodiments are all data authorized by the user for use.

[0059] The beneficial effects of this embodiment are:

[0060] The association results are obtained through multiple models, and the association credibility of family accounts and personal accounts is verified based on the model confidence. On the one hand, the integrity of the association results is improved through the multi-angle data of multiple models, and on the other hand, the accuracy of the association results is improved through cross-validation integration based on association credibility.

[0061] Based on the above embodiment, this embodiment provides a method for calculating model confidence and association credibility, which is described in detail as follows.

[0062] The recognition model includes a baseline recognition model and a comparative recognition model; the model confidence of the baseline recognition model is a set baseline confidence; the model confidence of the comparative recognition model is a relative confidence obtained based on the association result set of the baseline recognition model, the association result set of the comparative recognition model and the baseline confidence.

[0063] The relative confidence C of the contrast recognition model satisfies:

[0064]

[0065] Where C0 is the baseline confidence level; n is the number of accounts in the first household; m is the number of accounts in the second household;

[0066] The first family account refers to a family account in the association result set of the comparative identification model that belongs to the association result set of the benchmark identification model; the second family account refers to the first family account including at least one correct personal account; the correct personal account refers to a personal account corresponding to the family account that is the same as the first family account in the association result set of the benchmark identification model.

[0067] The step of calculating the association credibility between the family account and the personal account corresponding to the family account based on the at least two groups of association results and the model confidence of the recognition model includes:

[0068] Based on the association result, determine whether the personal accounts corresponding to the family account include the target personal account:

[0069] If the target personal account is included, the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result;

[0070] If the target personal account is not included and the family account does not have a corresponding personal account, the credibility factor of the association result is set to zero;

[0071] If the target personal account is not included, and the family account has a corresponding personal account, the inverse of the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result;

[0072] The credibility factors of at least two groups of association results are superimposed to obtain the association credibility between the family account and the target personal account.

[0073] The beneficial effects of this embodiment are:

[0074] The association results are obtained through multiple models, and the association credibility of family accounts and personal accounts is verified based on the model confidence. On the one hand, the integrity of the association results is improved through the multi-angle data of multiple models, and on the other hand, the accuracy of the association results is improved through cross-validation integration based on association credibility.

[0075] Based on any of the above embodiments, this embodiment provides a specific solution for identifying the model as follows.

[0076] The identification model includes any one or a combination of any two of a resident location elimination model based on business support data, a personal account mining model based on family account communications, a user relationship mining model based on application user identification, a relationship identification model based on business complaint data, a relationship identification model based on installation and maintenance data, and a relationship identification model based on account service application data.

[0077] The resident location elimination model based on the business support data is a model based on the relationship data between family accounts and personal accounts, the nighttime resident location of the personal account, and the range of the resident network cell of the family account in the business support data, and eliminates the association results in the relationship data between family accounts and personal accounts in the business support data where the nighttime resident location of the personal account is not within the range of the resident network cell of the family account;

[0078] The nighttime permanent location of the personal account is obtained through control plane signaling.

[0079] The personal account mining model based on family account communication is a model that obtains the device identification code of the communication device based on the family account communication data, and obtains the personal account used by the communication device according to the device identification code.

[0080] The user relationship mining model based on application user identification is a model for acquiring family accounts and personal accounts that communicate using the same user identification based on application user identification data.

[0081] The beneficial effects of this embodiment are:

[0082] The association results are obtained through multiple models, and the association credibility of family accounts and personal accounts is verified based on the model confidence. On the one hand, the integrity of the association results is improved through the multi-angle data of multiple models, and on the other hand, the accuracy of the association results is improved through cross-validation integration based on association credibility.

[0083] Based on any of the above embodiments, this embodiment provides an AI-based six-model method for identifying the relationship between family and personal accounts in combination with specific application scenarios.

[0084] In order to solve the problem that the existing methods for identifying the relationship between home and personal communication accounts are single, inaccurate and incomplete, this proposal proposes six different identification models that can independently realize the identification of the relationship between home and personal communication accounts; the accuracy and completeness of the identification of different models are compared, and operators can flexibly combine them according to the actual situation in the province; a method for evaluating the credibility of the model and a method for integrating the identification results of multiple models are given, especially the method for dealing with inconsistencies in the results of multiple methods.

[0085] Since the six models are based on different business principles and data sources and perform identification from different perspectives, the identification results are relatively independent. Through cross-validation and credibility judgment, the integrated results can effectively improve the accuracy and completeness of identification.

[0086] The six family and personal communication account relationship identification models are as follows:

[0087] 1. Eliminate abnormal relationships based on the mobile user's permanent location based on business support data, i.e., a permanent location elimination model based on business support data;

[0088] 2. Mobile phone number mining based on home broadband xDR, i.e., a personal account mining model based on family account communications;

[0089] 3. User relationship mining based on third-party APP user identification, that is, user relationship mining model based on application user identification;

[0090] 4. User-home broadband relationship identification based on home broadband complaints, i.e., a relationship identification model based on service complaint data;

[0091] 5. Automatic identification based on on-site installation and maintenance personnel (installation and maintenance APP), that is, a relationship recognition model based on installation and maintenance data;

[0092] 6. Relationship identification based on user self-service (APP), that is, a relationship identification model based on account service application data.

[0093] The six recognition models will be introduced in detail below.

[0094] 1. Eliminate abnormal relationships based on mobile users’ permanent locations based on industry data

[0095] Based on the existing data on the relationship between home broadband and mobile accounts, the nighttime location of the mobile phone number is compared with the boundary of the residential network cell to which the home broadband account belongs, and the association between the mobile phone's permanent location and the residential network cell location that does not match is eliminated.

[0096] Specific methods:

[0097] 1) Obtain analysis of the mobile phone’s nighttime location.

[0098] Because user-plane internet traffic flows over the home broadband network after a mobile phone accesses the home broadband via Wi-Fi, the control plane S1-MME signaling xDR is used. Based on the user's nighttime cell dwell time, the user's nighttime resident cell is determined. This is then correlated with resource engineering parameter data to obtain the user's nighttime resident location's latitude and longitude. This is standardized to a GIS point format.

[0099] 2) Obtain the range of the home broadband residential network area.

[0100] Obtain the boundary information of the residential network community to which the broadband account belongs based on the comprehensive information data.

[0101] Normalized to GIS polygon format.

[0102] Since the mobile phone location is approximated using the cell's longitude and latitude, the boundaries can be expanded, such as by using the GIS function fixed distance buffer.

[0103] 3) The mobile phone’s nighttime location matches the home broadband network area.

[0104] Based on the business and expense account relationship pair, determine whether the permanent location is within the boundary range. If so, retain the relationship; otherwise, remove it.

[0105] To determine the relationship between points and surfaces, the ray method can be used through a program, or through GIS functions such as st_within().

[0106] 2. Mobile phone number mining based on home broadband xDR

[0107] Mobile APKs can obtain phone-related permissions, including the ability to read device identifiers such as IMEI, MSISDN, and IMSI. This information is reported in certain app interactions. When a user uses their phone to access the internet via home broadband and the app reports this information, data mining of the home broadband HTTP xDR can extract the phone identifier and associate it with the home broadband account.

[0108] Specific methods:

[0109] 1) Establish a mobile phone identification mining rule base based on mobile xDR data

[0110] a. Based on current operator xDR data specifications and information content, mining this information requires the following conditions:

[0111] - Information is reported via HTTP (HTTPS, i.e. parsing of encrypted HTTP headers, is not supported)

[0112] -Information is reported via the query parameter in the URL (POST form data recording is not supported)

[0113] b. Analyze mobile network HTTP xDR data and establish candidate mining rules. Perform regular expression matching on the parameters following the "?" field in the URL, using the K=V format, where K is "IMSI," "IMEI," or "MSISDN," and V is a fixed-length number. For example, IMSI is "^4600[0-9]{11}" and MSISDN is "^(86){0,1}139[0-9]{8}."

[0114] c. Cross-validate and filter valid rules. For xDRs matching the candidate rules, extract the K and V fields and compare them with the IMSI, IMEI, or MSISDN fields in the mobile public fields of the xDR. If they match, the reported mobile phone identifier is valid, and the rule is valid. For the IMEI, considering dual-SIM cards may have dual IMEIs, only the first 8 digits of the IMEI, namely the TAC portion, are compared.

[0115] d. Expand the fields of the valid rules to establish a mobile phone identification mining rule base. From the xDR matching the valid rules in step c, extract the business category, business subcategory, and Host fields, along with the K and V regular expressions, to generate mining rules. These multiple rules constitute the mobile phone identification mining rule base.

[0116] 2) Using the rule base to mine mobile phone identifiers for home broadband xDR data

[0117] a. Processing home broadband HTTP xDR, according to the above rule base, matching and mining the above mobile phone user identifier, such as K = "IMSI", V = a string of numbers starting with 460.

[0118] b. For the xDR that extracts the mobile user ID, obtain the home broadband account. For home broadband users with non-public IP addresses, correlate the NAT logs with the "User IP" field in the xDR and then retrieve the private IP address. Then, correlate the private IP address with the Radius logs and retrieve the home broadband account.

[0119] 3) Mobile phone account standardization

[0120] The mobile phone number mined according to the MSISDN rule is directly used as the mobile phone account.

[0121] According to the IMSI or IMEI representation mined according to the IMSI or IMEI rules, a mapping relationship among IMSI, IMEI, and MSISDN is established through mobile xDR, and the IMSI or IMEI is converted into MSISDN.

[0122] Eliminate IoT private network numbers with an MSISDN length of 13 digits (excluding 86).

[0123] Standardize the mobile phone number format, remove the prefix 86, and output the relationship between home broadband account and mobile phone account.

[0124] 3. User relationship mining based on APP user identification

[0125] Apps, such as social, video, and gaming applications, have unique user identifiers. Some embedded third-party SDKs for user behavior analysis also have cross-app unique user identifiers. Some applications offer both mobile and desktop access and use the same unique user identifier. These user identifiers are collectively referred to as OTT accounts. XDR data analysis can identify the OTT accounts transmitted by apps during specific interactions.

[0126] By establishing OTT account identification rules, OTT accounts are mined for mobile and home broadband xDRs respectively, and mobile and home broadband accounts are associated through the same OTT account.

[0127] Specific methods:

[0128] 1) Establish an OTT account rule library

[0129] a. Establish an alternative rule base

[0130] Based on mobile xDR, candidate rules are extracted for URL parameter parts and cookies in the form of K=V.

[0131] Where K is a keyword. Different apps have different interface designs and implementations, so K is completely configurable. Based on packet capture data (with tags) and xDR data (without tags), machine learning training is used to obtain common keywords such as uid, uuid, device_id, and device_code.

[0132] V restricts the format of the identifier, such as decimal or hexadecimal number composition, and the length of the identifier.

[0133] Extract the regular expressions of business categories, business subcategories, Host, K, and V to generate an alternative rule base.

[0134] b. Conflict Detection

[0135] Through long-term testing and conflict detection, non-permanent identifiers such as session identifiers, identifiers of accessed resources, and other non-unique identifiers are eliminated from alternative rules.

[0136] Based on the candidate rule base, mobile xDR data over a long period of time (e.g., one week) is identified. For the same user and the same rule, rules with different identifiers are extracted and removed.

[0137] c. Rule merging

[0138] Merge different business categories, hosts, and KV rules based on the same user and identifier. This is applicable to third-party SDKs or different apps within the same system that use the same identifier.

[0139] 2) Mining of OTT accounts for mobile and home broadband xDR.

[0140] Based on the rule base, OTT account mining is performed on mobile and home broadband HTTP xDR respectively.

[0141] Among them, the home broadband xDR association is backfilled with the home broadband account.

[0142] Enter the mobile or home broadband account and OTT account respectively.

[0143] 3) Mobile and home broadband account association.

[0144] For the results of step 3), the mobile and home broadband account relationships are generated based on the same OTT account, and then output after standardization.

[0145] 4. Identifying User-Home Broadband Relationships Based on Home Broadband Complaints

[0146] Based on the home broadband account and mobile number in the home broadband general complaint record, identify and establish the association between accounts.

[0147] 5. Automatic identification based on on-site installation and maintenance personnel (installation and maintenance APP)

[0148] During home broadband installation and maintenance visits, face-to-face interviews are conducted to record the mobile phone numbers of home broadband users. This process can be embedded in the installation and maintenance app to streamline data collection and reporting.

[0149] 6. Relationship identification based on user self-service (APP)

[0150] Through the self-service APP downloaded and installed by the user, the user's mobile phone account information and the actual home broadband information used are collected to obtain the correlation between the two.

[0151] The following will explain the feature analysis and expected results comparison of these six recognition models.

[0152] The above models are based on different business principles and data sources, and their recognition results have different characteristics. The analysis is as follows:

[0153] - Eliminate abnormal relationships based on the mobile user's permanent location based on industry support data.

[0154] This method covers all users of the original data of industry support data. After eliminating abnormal relationships, the data integrity is reduced while the data accuracy is improved. It can be used as a benchmark data for multi-modal comprehensive analysis.

[0155] -Mobile phone number mining based on home broadband xDR

[0156] This method can easily establish a rule base through mining and comparison of mobile xDR while ensuring the accuracy of recognition.

[0157] Due to limitations such as HTTPS encrypted transmission and xDR not supporting recording of POST content, data integrity in terms of user coverage is relatively low.

[0158] After the rules are established, only home broadband xDR mining and analysis is required, and deployment and implementation are relatively easy.

[0159] Supports identification of multiple individual users under one home broadband account.

[0160] Supports identification of mobile phone numbers from different networks.

[0161] -User relationship mining based on third-party APP user identification

[0162] Due to the richness of apps and the constant emergence of new ones, traditional manual analysis and mining methods are difficult to cope with. This method uses AI technology to discover alternative recognition rules and ensures recognition accuracy through a conflict detection mechanism.

[0163] This method requires mining both mobile and home broadband data simultaneously. Since users access the Internet through mobile phones and home broadband at different times, it is necessary to associate the OTT account identification result data with at least daily granularity.

[0164] Supports identification of multiple individual users under one home broadband account.

[0165] - User-home broadband relationship identification based on home broadband complaints, automatic identification based on on-site installation and maintenance personnel (installation and maintenance APP), and association relationship identification based on user self-service (APP)

[0166] These three methods are suitable for specific scenarios and users, and have relatively low data integrity, so they can be used as supplementary means.

[0167] The following describes the integration method of multi-model recognition results.

[0168] Establish a method for evaluating the credibility of recognition results, including the credibility of single-model and multi-model recognition results.

[0169] The optimal recognition result of multiple models is selected based on model credibility.

[0170] 1) Select a baseline model.

[0171] Aggregate multiple model recognition results, namely, models, home broadband-mobile account pairs.

[0172] Filter account pairs that yield the same results from different models, i.e., home broadband-mobile account pairs that appear ≥ 2 times.

[0173] Count the number of times these accounts are recognized by different models, and take the model with the most recognition times as the baseline model.

[0174] 2) Relative credibility of a single model.

[0175] a. The relative credibility of the benchmark model is C=1.

[0176] b. Compare other models (comparison models) with the benchmark model one by one and calculate the relative credibility of the comparison models.

[0177] The number of home broadband accounts that are recognized by both the baseline model and the comparison model is n. The number of mobile accounts that are recognized is m, and the number of those that are recognized is marked as trustworthy. The credibility of the comparison model is C = m / n.

[0178] c. For models 2 and 3 as comparison models, since one home broadband account may identify multiple mobile accounts, any one of the mobile accounts that matches is considered credible.

[0179] d. Ideally, the credibility of models 2 and 3 should be evaluated separately according to different rules.

[0180] e. Note that as data accumulates and rules change (such as Models 2 and 3), the credibility should be updated accordingly based on the new data results.

[0181] 3) Integration of multi-model results

[0182] Assume that for a home broadband account, there are six models A, B, C, D, E, and F, which each identify a different number of mobile phone accounts.

[0183] Group the mobile phone numbers and summarize the model credibility to obtain the number credibility. Add / subtract the highest number credibility from the other groups. If the result is greater than 0.5, output the result and represent the relationship credibility.

[0184] For example:

[0185]

[0186]

[0187]

[0188]

[0189]

[0190] The beneficial effects of this embodiment are:

[0191] The data accuracy problem of the existing technology is solved by permanent position comparison;

[0192] Five other recognition models were proposed to improve data integrity;

[0193] A method for evaluating model credibility is established, and a method for integrating multiple model identification results based on model credibility is given;

[0194] The following describes the multi-model-based family and personal account relationship identification device provided by the present invention. The multi-model-based family and personal account relationship identification device described below and the multi-model-based family and personal account relationship identification method described above can be referenced to each other.

[0195] An embodiment of the present invention provides a multi-model-based system for identifying relationships between household and personal accounts, including:

[0196] Association module 1 is configured to obtain at least two sets of association results for the same family account based on different recognition models; the association results are included in a set of association results, including the family account and the individual accounts corresponding to the family account; the association result set is obtained by the recognition model based on the input data set, and includes the correspondence between multiple family accounts and multiple individual accounts;

[0197] Trust module 2, configured to calculate the association credibility between the family account and the personal account corresponding to the family account based on the at least two sets of association results and the model confidence of the recognition model;

[0198] Conclusion module 3 is used to add the family accounts whose association credibility meets the set range and the personal accounts corresponding to the family accounts to the identification conclusion.

[0199] Furthermore, the trusted module 2 includes:

[0200] A judging unit is configured to judge, based on the association result, whether the personal accounts corresponding to the family account include a target personal account:

[0201] If the target personal account is included, the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result;

[0202] If the target personal account is not included and the family account does not have a corresponding personal account, the credibility factor of the association result is set to zero;

[0203] If the target personal account is not included, and the family account has a corresponding personal account, the inverse of the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result;

[0204] The superposition unit is used to superimpose the credibility factors of at least two groups of association results to obtain the association credibility between the family account and the target personal account.

[0205] The beneficial effects of this embodiment are:

[0206] The association results are obtained through multiple models, and the association credibility of family accounts and personal accounts is verified based on the model confidence. On the one hand, the integrity of the association results is improved through the multi-angle data of multiple models, and on the other hand, the accuracy of the association results is improved through cross-validation integration based on association credibility.

[0207] Figure 3 An example of a physical structure diagram of an electronic device is shown below. Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communications bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communications bus 340. The processor 310 may invoke logic instructions in the memory 330 to execute a multi-model-based method for identifying relationships between family and personal accounts. The method includes: obtaining at least two sets of association results for the same family account based on different recognition models; the association results are results in a set of association results that include the family account and the personal accounts corresponding to the family account; the association result set is a set of corresponding relationships between multiple family accounts and multiple personal accounts obtained by the recognition model based on an input data set; calculating the association credibility between the family account and the personal accounts corresponding to the family account based on the at least two sets of association results and the model confidence of the recognition model; and adding the family account and the personal accounts corresponding to the family account whose association credibility meets a set range to the recognition conclusion.

[0208] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0209] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the multi-model-based family and personal account relationship identification method provided by the above methods, the method including: for the same family account, obtaining at least two sets of association results based on different identification models; the association results are in the association result set, including the results of the family account and the personal account corresponding to the family account; the association result set is obtained by the identification model according to the input data set, including a set of corresponding relationships between multiple family accounts and multiple personal accounts; based on the at least two sets of association results and the model confidence of the identification model, the association credibility of the family account and the personal account corresponding to the family account is calculated; the family account and the personal account corresponding to the family account whose association credibility meets the set range are added to the identification conclusion.

[0210] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the multi-model-based family and personal account relationship identification method provided by the above-mentioned methods, the method comprising: obtaining at least two sets of association results for the same family account based on different identification models; the association results are in an association result set, including the results of the family account and the personal account corresponding to the family account; the association result set is obtained by the identification model based on an input data set, including a set of corresponding relationships between multiple family accounts and multiple personal accounts; the association credibility of the family account and the personal account corresponding to the family account is calculated based on the at least two sets of association results and the model confidence of the identification model; the family account and the personal account corresponding to the family account whose association credibility meets the set range are added to the identification conclusion.

[0211] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0212] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0213] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A multi-model based method for identifying relationships between household and personal accounts, characterized in that: include: For the same family account, obtain at least two sets of association results based on different recognition models; The association result is a result of the family account and the personal account corresponding to the family account in the association result set; the association result set is a set of correspondences between multiple family accounts and multiple personal accounts obtained by the recognition model based on the input data set; Calculating, based on the at least two sets of association results and the model confidence of the recognition model, an association credibility between the family account and the personal account corresponding to the family account; Adding the family account whose association credibility meets the set range and the personal account corresponding to the family account to the identification conclusion; The recognition model includes a reference recognition model and a comparison recognition model; The model confidence of the benchmark recognition model is a set benchmark confidence; The model confidence of the comparative recognition model is a relative confidence obtained based on the association result set of the benchmark recognition model, the association result set of the comparative recognition model, and the benchmark confidence; The relative confidence C of the contrast recognition model satisfies: Where C0 is the baseline confidence level; n is the number of accounts in the first household; m is the number of accounts in the second household; The first family account refers to a family account in the association result set of the comparative recognition model that belongs to the association result set of the benchmark recognition model; the second family account refers to the first family account that includes at least one correct personal account; The correct personal account refers to the personal account corresponding to the same family account as the first family account in the association result set of the benchmark recognition model; The step of calculating the association credibility between the family account and the personal account corresponding to the family account based on the at least two groups of association results and the model confidence of the recognition model includes: Based on the association result, determine whether the personal accounts corresponding to the family account include the target personal account: If the target personal account is included, the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result; If the target personal account is not included and the family account does not have a corresponding personal account, the credibility factor of the association result is set to zero; If the target personal account is not included, and the family account has a corresponding personal account, the inverse of the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result; Superimposing the credibility factors of at least two sets of association results to obtain the association credibility between the family account and the target personal account; The recognition model includes any one or a combination of any two or more of a resident location elimination model based on business support data, a personal account mining model based on family account communications, a user relationship mining model based on application user identification, a relationship recognition model based on business complaint data, a relationship recognition model based on installation and maintenance data, and a relationship recognition model based on account service application data; the benchmark recognition model is obtained by selecting one of the recognition models; the comparative recognition model is other recognition models in the recognition model except the benchmark recognition model.

2. The multi-model based household and personal account relationship identification method according to claim 1, characterized in that: The resident location elimination model based on the business support data is a model based on the relationship data between family accounts and personal accounts, the nighttime resident location of the personal account, and the range of the resident network cell of the family account in the business support data, and eliminates the association results in the relationship data between family accounts and personal accounts in the business support data where the nighttime resident location of the personal account is not within the range of the resident network cell of the family account; The nighttime permanent location of the personal account is obtained through control plane signaling.

3. The multi-model based household and personal account relationship identification method according to claim 1, characterized in that: The personal account mining model based on family account communication is a model that obtains the device identification code of the communication device based on the family account communication data, and obtains the personal account used by the communication device according to the device identification code.

4. The multi-model based household and personal account relationship identification method according to claim 1, characterized in that: The user relationship mining model based on application user identification is a model for acquiring family accounts and personal accounts that communicate using the same user identification based on application user identification data.

5. A multi-model based family and personal account relationship identification system, characterized by: include: an association module, configured to obtain at least two sets of association results based on different recognition models for the same family account; The association result is a result of the family account and the personal account corresponding to the family account in the association result set; the association result set is a set of correspondences between multiple family accounts and multiple personal accounts obtained by the recognition model based on the input data set; a trust module, configured to calculate, based on the at least two sets of association results and the model confidence of the recognition model, an association credibility between the family account and the personal account corresponding to the family account; A conclusion module, configured to add the family accounts whose association credibility meets a set range and the personal accounts corresponding to the family accounts to the identification conclusion; The recognition model includes a reference recognition model and a comparison recognition model; The model confidence of the benchmark recognition model is a set benchmark confidence; The model confidence of the comparative recognition model is a relative confidence obtained based on the association result set of the benchmark recognition model, the association result set of the comparative recognition model, and the benchmark confidence; The relative confidence C of the contrast recognition model satisfies: Where C0 is the baseline confidence level; n is the number of accounts in the first household; m is the number of accounts in the second household; The first family account refers to a family account in the association result set of the comparative recognition model that belongs to the association result set of the benchmark recognition model; the second family account refers to the first family account that includes at least one correct personal account; The correct personal account refers to the personal account corresponding to the same family account as the first family account in the association result set of the benchmark recognition model; The trusted module is specifically used to: Based on the association result, determine whether the personal accounts corresponding to the family account include the target personal account: If the target personal account is included, the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result; If the target personal account is not included and the family account does not have a corresponding personal account, the credibility factor of the association result is set to zero; If the target personal account is not included, and the family account has a corresponding personal account, the inverse of the model confidence of the recognition model to which the association result belongs is used as the credibility factor of the association result; Superimposing the credibility factors of at least two sets of association results to obtain the association credibility between the family account and the target personal account; The identification model includes any one or a combination of a resident location elimination model based on business support data, a personal account mining model based on family account communications, a user relationship mining model based on application user identification, a relationship identification model based on business complaint data, a relationship identification model based on installation and maintenance data, and a relationship identification model based on account service application data; The reference recognition model is obtained by selecting one of the recognition models; the comparison recognition model is the other recognition model in the recognition model except the reference recognition model.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the multi-model-based method for identifying the relationship between household and personal accounts are implemented as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-model-based household and personal account relationship identification method as described in any one of claims 1 to 4 are implemented.

8. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the multi-model-based household and personal account relationship identification method as described in any one of claims 1 to 4 are implemented.

Images